Log In

Arize Phoenix developed by Arize AI is an open-source AI observability library that enables comprehensive tracing and monitoring for your AI applications. By integrating Arize's observability tools with your Groq-powered applications, you can gain deep insights into your LLM worklflow's performance and behavior with features including:

  • Automatic Tracing: Capture detailed metrics about LLM calls, including latency, token usage, and exceptions
  • Real-time Monitoring: Track application performance and identify bottlenecks in production
  • Evaluation Framework: Utilize pre-built templates to assess LLM performance
  • Prompt Management: Easily iterate on prompts and test changes against your data

Python Quick Start (3 minutes to hello world)

1. Install the required packages:

pip install arize-phoenix-otel openinference-instrumentation-groq groq

2. Sign up for an Arize Phoenix account.

2. Configure your Groq and Arize Phoenix API keys:

export GROQ_API_KEY="your-groq-api-key"
export PHOENIX_API_KEY="your-phoenix-api-key"

3. (Optional) Create a new project or use the "default" project as your project_name below.

4. Create your first traced Groq application:

In Arize Phoenix, traces capture the complete journey of an LLM request through your application, while spans represent individual operations within that trace. The instrumentation automatically captures important metrics and metadata.

import os
from phoenix.otel import register
from openinference.instrumentation.groq import GroqInstrumentor
from groq import Groq

# Configure environment variables for Phoenix
os.environ["OTEL_EXPORTER_OTLP_HEADERS"] = f"api_key={os.getenv('PHOENIX_API_KEY')}"
os.environ["PHOENIX_CLIENT_HEADERS"] = f"api_key={os.getenv('PHOENIX_API_KEY')}"
os.environ["PHOENIX_COLLECTOR_ENDPOINT"] = "https://app.phoenix.arize.com"

# Configure Phoenix tracer
tracer_provider = register(
    project_name="default",
    endpoint="https://app.phoenix.arize.com/v1/traces",
)

# Initialize Groq instrumentation
GroqInstrumentor().instrument(tracer_provider=tracer_provider)

# Create Groq client
client = Groq(api_key=os.getenv("GROQ_API_KEY"))

# Make an instrumented LLM call
chat_completion = client.chat.completions.create(
    messages=[{
        "role": "user",
        "content": "Explain the importance of AI observability"
    }],
    model="llama-3.3-70b-versatile",
)

print(chat_completion.choices[0].message.content)

Running the above code will create an automatically instrumented Groq application! The traces will be available in your Phoenix dashboard within the default project, showing detailed information about:

  • Application Latency: Identify slow components and bottlenecks
  • Token Usage: Track token consumption across different operations
  • Runtime Exceptions: Capture and analyze errors and rate limits
  • LLM Parameters: Monitor temperature, system prompts, and other settings
  • Response Analysis: Examine LLM outputs and their characteristics

Challenge: Update an existing Groq-powered application you've built to add Arize Phoenix tracing!

For more detailed documentation and resources on building observable LLM applications with Groq and Arize, see: